This is a repository copy of Adiposity and binge eating are related to liking and wanting for food in Brazil: A cultural adaptation of the Leeds Food Preference Questionnaire.
White Rose Research Online URL for this paper: http://eprints.whiterose.ac.uk/147174/
Version: Accepted Version
Article: de Carvalho-Ferreira, JP, Finlayson, G orcid.org/0000-0002-5620-2256, da Cunha, DT et al. (3 more authors) (2019) Adiposity and binge eating are related to liking and wanting for food in Brazil: A cultural adaptation of the Leeds Food Preference Questionnaire. Appetite, 133. pp. 174-183. ISSN 0195-6663 https://doi.org/10.1016/j.appet.2018.10.034
© 2018 Elsevier Ltd. All rights reserved. Licensed under the Creative Commons Attribution-Non Commercial No Derivatives 4.0 International License (https://creativecommons.org/licenses/by-nc-nd/4.0/).
Reuse This article is distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs (CC BY-NC-ND) licence. This licence only allows you to download this work and share it with others as long as you credit the authors, but you can’t change the article in any way or use it commercially. More information and the full terms of the licence here: https://creativecommons.org/licenses/
Takedown If you consider content in White Rose Research Online to be in breach of UK law, please notify us by emailing [email protected] including the URL of the record and the reason for the withdrawal request.
[email protected] https://eprints.whiterose.ac.uk/ 1 ADIPOSITY AND BINGE EATING ARE RELATED TO LIKING AND WANTING FOR 2 FOOD IN BRAZIL: A CULTURAL ADAPTATION OF THE LEEDS FOOD PREFERENCE 3 QUESTIONNAIRE
4 Joana Pereira de Carvalho-Ferreira1,2, Graham Finlayson3, Diogo Thimoteo da Cunha4, Gabriele 5 Caldas2, Daniel Bandoni5, Veridiana Vera de Rosso1
6 1Department of Biosciences, Federal University of São Paulo, Santos-SP, Brazil 7 2Post Graduation Program on Foods, Nutrition and Health, Federal University of São Paulo, 8 Santos-SP, Brazil 9 3Institute of Psychological Sciences, University of Leeds, Leeds, West Yorkshire, United Kingdom 10 4 School of Applied Sciences, Campinas State University, Limeira-SP, Brazil 11 5Department of Health, Clinic and Institutions, Federal University of São Paulo, Santos-SP, Brazil 12
13 Declarations of interest: none.
14 Corresponding authors: Joana Pereira de Carvalho-Ferreira and Veridiana Vera de Rosso, 15 Department of Biosciences, Federal University of São Paulo, Rua Silva Jardim 136, Santos, SP CEP 16 11015-020, Brazil. Email: [email protected] (Carvalho-Ferreira JP) and 17 [email protected] (de Rosso VV). 18
19
20
21
22
23
24
25
26 27 ABSTRACT
28 The Leeds Food Preference Questionnaire (LFPQ) measures separable psychological
29 components of food reward (Liking and Wanting). In this study a cultural adaptation of the LFPQ
30 for a Brazilian population (LFPQ-BR) was examined by comparing liking and wanting scores in
31 fasted and fed states and their association with adiposity and disturbed eating. A culturally
32 adapted food picture database was validated by an online questionnaire completed by 162
33 individuals. Cluster analysis verified if the foods were accurately perceived in terms of
34 sweetness, fat and calorie content. Subsequently, 48 male (N=21) and female (N=27) adults with
35 mean Body Mass Index 26.6 (0.9) kg/m2, and mean age 32.8 (1.4) years, were evaluated by the
36 LFPQ-BR before and after a fixed test meal. The Binge Eating Scale was used to measure binge
37 eating symptoms. There was a decrease in explicit liking, implicit wanting, and explicit wanting
38 scores for food in general in the fed condition. The implicit and explicit wanting and explicit liking
39 scores for high-and-low fat savoury food decreased and for high-and-low fat sweet foods
40 increased to a greater extent after the savoury test meal. Body Mass Index was found to predict
41 implicit wanting for high fat relative to low fat foods. Binge eating symptoms predicted high fat
42 sweet explicit liking and explicit wanting in the fed condition. Finally, high fat sweet preference
43 was found to be sex-related as females had greater implicit wanting for high fat sweet foods in
44 fasted and fed states. The results presented here indicate that the LFPQ-BR is a useful
45 instrument for the evaluation of liking and wanting for food in Brazil.
46 Keywords: food reward; liking; wanting; binge eating; adiposity; Brazil.
47
48
49 50 INTRODUCTION
51 Overeating is an important risk factor for weight gain and because obesity prevalence is
52 increasing worldwide (Swinburn et al., 2011), acquiring a comprehensive view of the factors that
53 lead to overeating is crucial. It is already known that eating behaviour is not only modulated by
54 hypothalamic circuitry, but is also determined by the hedonic system responding to an
55 obesogenic environment (Berthoud, 2006). Natural rewards, like food, stimulate activation in
56 the mesolimbic dopaminergic pathway (Lutter and Nestler, 2009) and individual variability in
57 sensitivity to reward is a psychobiological trait linked to the development of obesity (Beaver,
58 2006).
59 Therefore, overeating is thought to be more than an imbalance of hormonal satiation
60 and satiety signalling, but also occurs due to an excessive or weakened response to the hedonic
61 aspects of food (Blundell and Finlayson, 2004). Considering the role of cognitive and hedonic
62 aspects of eating behaviour helps to understand more than meal size and frequency, but also
63 food preferences and choice, which are partly driven by the motivation and experience of
64 pleasure obtained from food (wanting and liking for food). This distinction is key to comprehend
65 overeating leading to weight gain in an environment where highly palatable and energy-dense
66 foods are plentiful and affordable (Dalton et al., 2013b). Importantly, it is suggested that high
67 palatable foods are consumed frequently even when energy needs are satisfied, while less tasty
68 foods are not overconsumed and this is one of the key risk factors associated with obesity
69 (Kenny, 2011). Thus, palatability is a key contributor to the decision to eat particularly for
70 individuals susceptible to reward-driven eating.
71 The Leeds Food Preference Questionnaire (LFPQ) developed by Finlayson et al. (2007)
72 is a computer-based platform designed to measure the separate constructs of liking and wanting
73 according to key dimensions of food (e.g. fat/protein content and taste). The questionnaire
74 cales and includes an 75 s of decisions between pairs of
76 foods. Previous research with the LFPQ has demonstrated that liking for food, i.e. the perceived
77 or expected pleasure value of the food, the appreciation of its sensory proprieties, or a judgment
78 of the degree of pleasure it elicits (Dalton and Finlayson, 2014) is greater in fasted than fed states
79 (Finlayson et al., 2008) and that liking for a recently eaten food decreases in a manner consistent
80 with sensory specific satiation (Griffioen-Roose et al., 2010). Wanting on the other hand, i.e. the
81 attraction that is triggered by the perception of a food or a food related cue in the environment
82 (Dalton and Finlayson, 2014) is also increased for food in general in a fasted state (Alkahtni et
83 al., 2016; Finlayson et al., 2008). However, it appears that wanting is more variable than liking
84 and may differ moment to moment depending on a number of factors such as hunger state,
85 time of day and/or the amount of attentional resources available (Dalton and Finlayson, 2014).
86 For instance, disordered eating patterns (Dalton et al., 2013a; Finlayson et al., 2012) and a state
87 of macronutrient imbalance (Griffioen-Roose et al., 2012) have been linked to increased wanting
88 for specific foods.
89 Because cultural issues play a major role in food choice, selection, and consumption,
90 cultural adaptation of LFQP may be necessary for use in Brazil . Studies have already shown that
91 food choices and their motivators have a strong ethnic and cultural relationship (Januszewska
92 et al., 2011; Prescott et al., 2002). As an example, a traditional Brazilian dietary pattern is defined
93 by the consumption of rice, beans, green vegetables, potato, lettuce, eggs, milk, and meat
94 (Marchioni et al., 2011). In contrast, people in the United Kingdom are likely to eat white bread,
95 butter, tea and sugar, cakes, puddings, ham, bacon, potatoes, and vegetables (Pryer, 2001).
96 Importantly, a previous research using the Leeds Food Preference Questionnaire (Leenaars et
97 al., 2016) indicated as a limitation of their study the imperfect suitability of using the translated
98 LFPQ for a Dutch population. They mentioned that some foods in the food database used in the
99 LFPQ would be less familiar to Dutch consumers in comparison with those from the UK. For
100 example, participants had difficult to identify one food as savoury or sweet. Therefore, 101 performing a validation study for the Brazilian Portuguese version (translation and food
102 database suitability) is of major importance in using the instrument in Brazil.
103 Brazil is currently facing an epidemic of obesity and overweight (Malta et al., 2014).
104 Recently, a food guide for the Brazilian population has been published suggesting that people
105 should consume less processed foods, i.e. foods high in fat, salt and sugar, indicating the harm
106 excessive consumption of these foods may bring about (Ministry of Health of Brazil, 2014).
107 However, the food choice goes beyond the perceived risk and benefits. There are several
108 benefits of having instruments to test components of food reward in a population. A reliable
109 method that allows the quantification of liking and wanting for different dimensions of food may
110 be valuable to link specific food preferences to health problems (such as weight gain and eating
111 disease), or identify risk factors (Dalton et al., 2013a; Finlayson et al., 2012). Moreover, it is
112 useful to test different types of diets under different contexts (Cameron et al., 2014; Griffioen-
113 Roose et al., 2012, 2011; Hopkins et al., 2016) and to link food reward to other circumstances,
114 such as sleep restriction (Leenaars et al., 2016).
115 The current study aimed to adapt the original LFPQ for the Brazilian population (LFPQ-
116 BR). Furthermore, we tested the sensitivity of this cultural adaptation by comparing liking and
117 wanting scores in fed and fasted states and their association with adiposity and scores on the
118 Binge Eating Scale.
119
120
121
122
123 124 MATERIAL AND METHODS
125 To meet the aims of the present study, some steps were performed. First, all the text from
126 the task was translated into Brazilian Portuguese and a food picture database was created and
127 validated for the Brazilian population. Following this cultural adaptation, an experimental
128 procedure was conducted to test the sensitivity of this new version of the LFPQ for a Brazilian
129 population.
130 a) Food database validation 131
132 To create a food database for the LFPQ-BR, ready-to-eat popular foods in Brazil were
133 photographed according to standardised procedure and 60 images went through an online
134 validation process. An online questionnaire was created and released by email and social media
135 and 162 respondents completed it. The questionnaire aimed to verify if the food pictures were
136 accurately recognized, habitually consumed, and that they were correctly perceived in terms of
137 fat content, number of calories and sweetness (high/low). The answers of any registered
138 dietician (i.e. nutritionist in Brazil) or nutrition specialists were excluded.
139 Each food picture was presented one at a time and the respondents were asked the
140 following questions for each one: do you recognize this food? (yes; I am not sure; no); have you
141 already eaten this food? (yes; I am not sure; no); do you like this food? (yes; more or less; no).
142 Additionally, participants were asked to rate on a 7-point Likert scale (with anchor points at
143
144 contain? How sweet is this food? How calorie-dense is this food?
145 Foods were considered adequate if: 80% of the sample recognize, habitually consume them,
146 and like them and if their mean values on the 7-point Likert scale answers were above 5 for the
147 high content of fat, calorie and sweet taste; and lower than 3 for the low content of fat, calorie
148 and sweet taste. Further, hierarchical W 149 distinctions among the food items. Three clusters were constructed: 1) content of fat; 2) content
150 of sweetness and 3) calorie amount, using the answers for each food on the Likert scale.
151 Dissimilarity dendograms, analysis of intraclass variance and centroid mean distance were
152 constructed.
153
154 b) The Leeds Food Preference Questionnaire (LFPQ) and Brazilian Portuguese Version of
155 the LFPQ (LFPQ-BR)
156 The LFPQ is a computer-based task developed to provide measures of different components
157 of food preference and food reward. The measures taken by the task and the methodology used
158 are summarized in Figure 01. Participants were presented with an array of 16 ready-to-eat food
159 images, which are common in the diet. Food images were chosen from a validated database,
160 four from each category, as follows: high fat sweet (HFSW), high fat savoury (HFSA), low fat
161 sweet (LFSW) and low fat savoury (LFSA) (Dalton and Finlayson, 2014; Finlayson et al., 2008).
162
163 Figure 1 Measures taken by the Leeds Food Preference Questionnaire. 164 Text from the original version of the LFPQ (Finlayson et al., 2007) was translated by one of
165 the authors, a Brazilian-Portuguese native speaker with English Skills, who translated the English
166 version into Brazilian-Portuguese. The original English task was also sent to a bilingual professor
167 who has expertise in the area and to a certified translator. The versions were compared,
168 discrepancies were discussed, and modifications were made to achieve the most accurate
169 Portuguese version of the task, which was piloted before the experiment.
170
171 Implicit wanting and food preference 172 173 Implicit wanting and food preference were measured using a forced choice
174 methodology. Participants are presented with 96 randomized food pairs ensuring every image
175 from each of the four categories is compared to every image from other categories, and are
176
177 ‘ s are covertly recorded and used to compute mean response times for
178 each food category. Both selection (positively contributing) and non-selection (negatively
179 contributing) are recorded to calculate implicit wanting scores (frequency-weighted algorithm).
180 A positive score indicates a more rapid preference for a given food category relative to the
181 alternatives in the task and a negative score indicates the opposite. A score of zero would
182 indicate that the category is equally preferred (Figure 2).
183
184 Figure 2 Representation of the paired foods instructions and the implicit wanting trials of the 185 LFPQ-BR.
186 *Which food do you most want to eat now?
187 188 Explicit Liking and Explicit Wanting 189 190 To measure explicit liking and explicit wanting, participants were presented with food
191 images individually and were
192 H
193 H
194 for the explicit wanting measure. Final scores range from 0 to 100 (Figure 3).
195
196 Figure 3 Representation of the single foods instructions, explicit liking (a) and explicit wanting 197 (b) trials of the LFPQ-BR.
198 (a)How pleasant would it be to taste some of this food now? 199 (b)How much do you want some of this food now? 200 * Not at all ------Extremely 201
202 Fat Appeal Bias and Taste Appeal Bias
203
204 For each one of the constructs of food reward (explicit liking and wanting; and implicit
205 wanting) the fat appeal bias and sweet appeal bias were calculated for explicit liking, explicit
206 wanting or implicit wanting, according to high fat relative to low fat foods (subtracting the mean 207 high fat values from the mean low fat values), and sweet relative to savoury foods (subtracting
208 the mean sweet values form the mean savoury values).
209 c) Experimental procedure using the LFPQ-BR 210 211 Participants and Study Design 212 213 Participants were recruited by social media and flyers with a call to participate in
214 research about food preferences and behaviour. Forty-eight adult (above 18 and under 60 years)
215 male and female participants were included. They participated in a voluntary basis, as they did
216 not receive any form of reward for the study. Exclusion criteria were smoking, pregnancy,
217 diagnosed metabolic disease such as diabetes and hypo - or hyperthyroidism, and allergies
218 and/or aversion to the food served in the test meal or included in the LFPQ-BR.
219 The experiment was set up at the Dietary Technique Laboratory, which has a private air-
220 conditioned room where the participants were evaluated. Screening was performed by email to
221 investigate inclusion and exclusion criteria, and participants were asked to attend the laboratory
222 following a 4-hour fast on a previously scheduled appointment between 11am and 13pm.
223 Informed consent was obtained, and height and weight were measured with participants
224 standing, wearing light clothes and no shoes. This was followed by testing the food preferences
225 of the participants in fasted state, evaluated by the LFPQ-BR. A standardised test meal was
226 provided, then 10 minutes after consuming it, the LFPQ-BR was undertaken again to evaluate
227 food preferences in the fed state. Subjective appetite measures were undertaken before and
228 after the test meal. Finally, the Binge Eating Scale was completed at the end of the experiment.
229 The whole visit for each participant lasted approximately 1 hour. This study was formally
230 approved by the ethical committee of the Federal University of São Paulo (#0531/2016) and was
231 carried out in accordance with the principles of the Declaration of Helsinki.
232 Test Meal 233 The test meal was fixed at the same amount for all participants and was piloted before
234 the experiment to adjust for taste acceptance and suitability of the amount provided. The lunch
235 consisted of 500 gram ( 650 kcal) of penne, meatballs and mixed buttered vegetable (carrots,
236 barred potato, broccoli and fine beans), 150 ml of orange juice and 150 ml of water. The test
237 meal was planned to be predominantly savoury and we aimed to plan a balanced meal in terms
238 of macronutrient distribution and calorie amount (Ministério do Trabalho e Emprego, 2006;
239 Ministry of Health of Brazil, 2014) (Table 1). It was prepared fresh every day and participants
240 were instructed to consume the meal in its entirety.
Table 1 Nutritional value of the test meal planned to be balanced in terms241 of macronutrient distribution and calorie amount. Food Total (g) Protein (g) Fat (g) Carbohydrates g) Pasta 120.5 4.08 0.48 27.60 Tomate sauce 110 1.54 10.01 7.40 Meat Balls 124 26.59 11.13 11.71 Buttered vegetebles 149 2.28 3.30 14.37 Orange juice 150 1.05 0.15 11.40 Total 653.5 35.54 25.07 72.48 242
243 Binge Eating Scale (BES) 244 245 Binge eating symptoms were assessed using the Binge Eating Scale (BES). This is a 16-
246 item self-report instrument to investigate behavioural manifestations (eight items) and feelings
247 and cognitions (eight items) of binge eating. For each item, three or four statements are given
248 that increase in severity and the participant is asked to select a statement that best describes
249 how they usually behave/feel. Scores are summed and binge eating behaviour is classified into
250 levels of severity: mild (scoring 17 or less), moderate (18- (Freitas et al.,
251 2001; Gormally et al., 1982).
252
253 Appetite Measures
254 255 In order to verify the efficacy of the manipulation of the hunger/satiation state,
256 subjective appetite measures were undertaken before and after the test meal, using a paper-
257 based visual analog scale (VAS), where the participant selected a position on a continuous
258 100mm linear scale to represent their answer to the question. To evaluate hunger the
259 participant answered the question
260 D
261 food could you eat? F
262 evaluate fullness t were
263
264
265 Statistical analysis
266 All variables underwent a compliance test to check their distribution against theoretical
267 curves. For this, histograms were analyzed and the Kolmogorov-Smirnov test for normality and
268 Levene test were applied to verify homoscedasticity.
269 General Linear Models (GLM) were used to observe the main effects and its interactions
270 of taste (savory or sweet), fat (high or low) and condition (fasted or fed) among dependent
271 variables: explicit liking, explicit wanting and implicit wanting. The t-Student test for related
272 samples was used to compare different scores between two dependent groups.
273 Multiple linear regression models were developed to investigate the relationship
274 between independent variables and dependent variables: explicit wanting of high fat sweet
275 food, explicit liking high fat sweet food, implicit wanting high fat sweet food, explicit wanting
276 sweet appeal bias and implicit wanting fat appeal bias. It was tested as independent variables
277 for P dependent variables and
278 only those with the statistically significant coefficient of regression remained. The insertion of 279 variables in the models was done by stepwise method with forward selection. The fit of the
280 models was evaluated by residual analysis.
281 All analyses were conducted using Statistical Package for Social Sciences (SPSS) for
282 Windows version 15.0.1. For all analyses, results were considered significant with p<0.05.
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299 300 RESULTS
301 a) Food database validation
302 Results from the validation demonstrated that 33 food images were properly recognized
303 and habitually consumed. Foods were considered adequate if their mean values on the 7-point
304 Likert scale answers were above 5 for the high content of fat, calorie and sweet taste; and lower
305 than 3 for the low content of fat, calorie and sweet taste. (Table 2). Further, we conducted
306 cluster analysis to confirm the classification of each food in the category they were assigned.
307
308
309
310
311
312
313
314
315
316
317
318
319 Table 2 Thirty three food cues validated for the LFPQ-BR, divided into the four categories.
Is it high Have Is it fat? Is it sweet? Food Reconize? Like? calorie? eaten? (mean ± SD) (mean ± SD) (mean ± SD) Chocolate M&Ms 100% 100% 90.2% 5.61 ± 0.8 5.94 ± 0.7 5.84 ± 0.7 Chocolate cheesecake 91.9% 90.2% 80.5% 5.45 ± 0.7 5.51 ±0.7 5.78 ± 0.8 Ice cream 100% 99.2% 90.2% 6.16 ± 0.8 6.00 ± 0.8 6.23 ± 0.8 Milk chocolate 99.2% 100% 91.1% 5.39 ± 0.8 5.54 ± 0.8 5.61 ± 0.8 Paçoca1 88.9% 92% 80.5% 5.30 ± 0.9 5.75 ± 0.9 5.67 ± 0.9
High Fat Sweet Sweet HighFat Chocolate chip cookies 99.2% 99.2% 80.5% 5.40 ± 0.8 5.52 ± 0.8 5.62 ± 0.9 Chocolate Balls* 99.2% 92% 80.5% 5.46 ± 0.9 5.54 ± 1.1 5.85 ± 0.8 Red Grapes 100% 100% 88% 1.41 ± 0.7 5.08 ± 0.7 2.46 ± 1.1 Fruit Salad 98.1% 97% 82.2% 1.80 ± 0.8 5.08 ± 0.7 2.73 ± 0.9 Watermelon 100% 98.1% 83.3% 1.00 ± 0.0 5.12 ± 0.7 2.22 ± 1.1 Papaya 100% 97% 82.2% 1.69 ± 0.8 5.08 ± 0.7 2.36 0.9 Melon 97% 97% 85.8% 1.41 ± 0.7 5.05 ± 0.7 1.99 ± 0.9
Low Fat Sweet Fat Low Banana 100% 99.2% 89.4% 1.58 ± 0.8 5.13 ± 0.7 2.77 ± 1.0 Mango* 92% 92% 85.8% 2.39 ± 1.2 5.20 ± 1.0 2.85 ± 1.1 Cheeseburger 100% 98.4% 80.5% 5.93 ± 0.8 1.40 ± 0.7 6.18 ± 0.7 Pepperoni pizza 99.2% 99.2% 90.2% 5.89 ± 0.9 1.77 1.0 6.15 ± 0.7 Coxinha2 99.4% 97.5% 85.8% 6.30 ± 0.7 1.64 ± 0.9 6.42 ± 0.6 Pastel frito3 98.1% 96.3% 83.3% 5.88 ± 0.7 1.77 ± 0.9 5.86 ± 0.7 Pão de queijo4 98.8% 98.1% 91.3% 5.08 ± 0.7 1.72 ± 0.9 5.09 ± 0.9 French Fries 98.4% 97.6% 86.2% 5.89 ± 0.8 1.26 0.5 5.98 ± 0.8
High Fat Savoury Savoury HighFat Croissant with ham and 100% 95.9% 86.2% 5.29 ± 1.1 1.93 ± 1.1 5.65 ± 1.0 cheese* Salami slices* 100% 96.7% 91.3% 5.98 ± 0.9 1.51 ± 1.1 5.86 ± 1.1 Peanuts* 95.6% 95.6% 85.1% 5.08 ± 1.1 2.07 ± 1.4 5.17 ± 1.2 Brocoli 100% 100% 82.1% 1.00 ± 0.0 1.00 ± 0.0 1.23 ± 0.5 Letuce-Tomato salad 100% 97.5% 95.1% 1.00 ± 0.0 1.35 ± 0.6 1.00 ± 0.0 Sweetcorn on the cob 100% 99.4% 87.7% 2.28 ± 1.0 1.91 ± 0.9 2.59 ± 0.9 Chicken-salad sandwich 97.5% 92% 80.2% 2.29 ± 0.7 1.69 ± 0.8 2.29 ± 0.9 Fine Beans 95.1% 90.2% 83.3% 1.64 ± 0.8 1.43 ± 0.8 1.89 ± 0.8 Tomatoes 100% 99.2% 83% 1.00 ± 0.0 2.03 ± 0.9 1.62 ± 0.8 Carrots* 100% 100% 85.8% 1.33 ± 0.6 2.37 ± 1.1 1.85 ± 1.0 Low Fat Savoury Fat Low Cod Meal* 99.2% 96.7% 82.9% 2.43 ± 1.1 2.43 ± 1.0 2.43 ± 1.1 Mixed salad* 100% 99.2% 80.5% 1.27 ± 0.6 1.35 ± 0.6 1.58 ± 0.8 Taboulli* 85.8% 82.2% 80.5% 2.38 ± 1.0 1.85 ± 1.1 2.85 ± 1.0 320 1 Candy made with peanut; 2 - Shredded chicken meat, covered in dough, molded into a shape resembling a chicken leg, battered and fried; 3 - half-circle or rectangle-shaped thin crust pies with assorted fillings, 321 fried in vegetable oil; 4 - a type of baked starch tart cookie, salt vegetable oil and cheese. SD = standard deviation. *Food image validated but not used in the present experiment. 322 Table 3 presents the results of the hierarchical cluster analysis for three variables
323 (sweetness, content of fat, and calories). For each variable, three clusters were generated (C1,
324 C2 and C3). 325 Table 3 Results of hierarchical cluster analysis for three variables (content of fat, content of sweetness and calorie amount). Mean Intraclass distance Suggested Variable Cluster Clustered food items variation of cluster name centroid French Fries, Cheeseburger, Pastel Frito, Pepperoni pizza, Coxinha, Pão de queijo, Chocolate cheesecake, Milk chocolate, Chocolate M&Ms, Paçoca, Ice cream, C1 94.96 9.37 High fat Chocolate chip cookies, Croissant with ham and cheese, Chocolate balls, Oreos, Cheese puffs, Glazed doughnut, Salami slices, Qindim, Peanuts. Chicken-salad sandwich, Letuce-Tomato salad, Tomatoes, Sweetcorn on the cob, Fine Beans, Brocoli, Content of Watermelon, Banana, Red Grapes, Papaya, Melon, C2 103.48 9.32 Low fat fat Fruit Salad, Carrots, Cod meal, Cucumber, Mixed salad, Tabouli, Mango, Apple, Strawberries, Orange, White rice. Cake, Biscuit cereal bar, Cream Crackers, Yogurt with strawberry and blueberry, White toast with butter, Bread roll, Caramel cereal bar, Cheese & crackers, Probably high C3 196.81 13.44 Beef sandwich, Muffin, Cashews, Scrambled eggs and fat toast, White toast with jam, Light yogurt, Brie, Marshmallow, Biscoito de polvilho, Boiled sweets. French Fries, Cheeseburger, Pastel Frito, Pepperoni pizza, Coxinha, Pão de queijo, Chicken-salad sandwich, Letuce-Tomato salad, Tomatoes, Sweetcorn on the cob, Fine Beans, Brocoli, Carrots, Croissant with ham C1 121.57 10.64 and cheese, Cod meal, Cucumber, Cheese puffs, Low sweet Sweetness Salami slices, Mixed salad, Peanuts, Tabouli, Cream Crackers, White toast with butter, Cheese & crackers, Beef sandwich, Cashews, Scrambled eggs and toast, Brie, Biscoito de polvilho, White rice. Chocolate cheesecake, Milk chocolate, Chocolate C2 104.63 9.84 High sweet M&Ms, Paçoca, Ice cream, Chocolate chip, cookies, Watermelon, Banana, Red Grapes, Papaya, Melon, Fruit Salad, Chocolate balls, Oreos, Glazed doughnut, Cake, Quindim, Mango, Caramel cereal bar, Muffin, Marshmallow, Boiled sweets. Biscuit cereal bar, Yogurt with strawberry and Probably high C3 170.35 11.99 blueberry, Apple, Bread roll, White toast with jam, Strawberries, Orange, Light yogurt. sweet French Fries, Cheeseburger, Pastel Frito, Pepperoni pizza, Coxinha, Pão de queijo, Chocolate cheesecake, Milk chocolate, Chocolate M&Ms, Paçoca, Ice cream, C1 102.73 9.70 Chocolate chip cookies, Croissant with ham and High calorie cheese, Chocolate balls, Oreos, Cheese puffs, Glazed doughnut, Salami slices, Cake, Quindim, Peanuts, Bread roll, Muffin, Marshmallow, Boiled sweets. Chicken-salad sandwich, Letuce-Tomato salad, Content of Tomatoes, Sweetcorn on the cob, Fine Beans, Brocoli, Calories C2 130.03 9.10 Watermelon, Banana, Red Grapes, Papaya, Melon, Low calorie Fruit Salad, Carrots, Cod meal, Cucumber, Mixed salad, Tabouli, Mango, Apple, Strawberries, Orange. Biscuit cereal bar, Cream Crackers, Yogurt with strawberry and blueberry, White toast with butter, Caramel cereal bar, Cheese & crackers, Beef sandwich, Probably high C3 176.35 12.68 Cashews, Scrambled eggs and toast, White toast with calorie jam, Light yogurt, Brie, Biscoito de polvilho, White rice.
326 327 The foods allocated on the C3 clusters of each analysis were not considered validated
328 because the perception of the total fat, sweetness and/or calorie content appeared confusing
329 to respondents. In other words, the third cluster indicated the participants did not consistently
330 perceive that food as high or low fat/sweet/calorie.
331 Considering the above, a set of 33 food images were validated (Supplementary Figure 1) and
332 24 food images were used in the present experiment, 4 for each category and 2 backups for each
333 category. The set of 16 foods were shown to participants before they started the task for the
334 first time and the backup images were used when the participant reported strong disliking for
335 any of the foods originally presented. The results of the food not validated for the LFPQ-BR can
336 be seen in Supplementary Table 1 and the food pictures not validated are shown in
337 Supplementary Figure 2.
338 b) LFPQ-BR experimental results
339 Twenty-one males and 27 females, born in Brazil, with a mean BMI 26.6 (±0.9) kg/m2,
340 ranging from 16.46 kg/m2 to 54.28 kg/m2 were evaluated in this study. Mean age was 32.8 (±1.4)
341 years and they presented mean weight of 75.1 (±2.9) kg and height 1.67 (±0.01) metres. Family
342 income was distributed as follow: 10.4% earn up to 2 minimum wages; 25% 2-4 minimum wages;
343 31% 5-10 minimum wages; 20.8% more than 10 minimum wages; and 12.8 % did not respond
344 to this question. Percent of educational attainment levels was: 2.1% lower secondary level; 4.2%
345 upper secondary level; 18.8% incomplete tertiary level; 20.8% complete tertiary level; and 54.1%
346 post graduate level.
347 The results are expressed as mean (standard error). Fasted and fed scores of explicit liking,
348 implicit wanting and explicit wanting of high fat savoury, low fat savoury, high fat sweet and low
349 fat sweet, fat appeal bias and taste appeal bias are shown in Table 4.
350 Table 4. Implicit wanting, explicit liking and explicit wanting for the food categories and appeal biases of the LFPQ- BR on fasted and fed states. Fasted Fed Explicit Implicit Explicit Implicit Explicit Explicit Liking Liking Wanting Wanting Wanting Wanting £ £ £ HFSA 66.8 (3.6) 16.3 (5.2) 60.9 (4.0) 9.5 (1.8) -47.8 (2.0) 8.8 (1.8) £ £ £ LFSA 65.6 (2.8) 15.1 (3.7) 66.4 (2.8) 13.6 (2.0) -26.2 (2.9) 12.2 (2.1) £ HFSW 46.3 (3.4) -24.0 (4.7) 41.3 (3.3) 57.4 (4.1) 38.7 (3.6) 53.2 (4.1) £ LFSW 51.6 (3.3) -7.5 (4.8) 50.4 (5.5) 52.3 (4.6) 35.2 (3.1) 49.1 (4.6) Fat Appeal Bias -2.0 (3.5) -7.6 (6.19) -7.3 (3.7) 0.5 (2.0) -9.0 (4.2) 0.3 (2.2) Sweet Appeal Bias -17.2 (3.5) -31.5 (5.7) -17.8 (3.7) 43.3 (3.8) 74.0 (2.8) 40.7 (3.7) HFSA = High Fat Savoury; LFSA = Low Fat Savoury; HFSW = High Fat Sweet; LFSW = Low Fat Sweet